Sense of Economic Gain from E-Commerce: Different Effects on Poor and Non-Poor Rural households
WangYu(王瑜) ...............................................................................................................................................
Abstract:
Sense of economic gain of e-commerce participation is an important aspect for evaluating the inclusiveness of e-commerce development. Based on the data of 6,242 rural households collected from the 2017 summer surveys conducted by the China Institute for Rural Studies (CIRS), Tsinghua University, this paper evaluates the effects of e-commerce participation on rural households’ sense of economic gain with the propensity score matching (PSM) method, and carries out grouped comparisons between poor and nonpoor households. Specifically, the “Self-evaluated income level relative to fellow villagers” measures respondents’ sense of economic gain in the relative sense, and “Percentage of expected household income growth (reduction) in 2018 over 2017” measures future income growth expectation. Findings suggest that e-commerce participation significantly increased sample households’ sense of economic gain relative to their fellow villagers and their future income growth expectation. Yet grouped comparisons offer different conclusions: E-commerce participation increased poor households’ sense of economic gain compared with fellow villagers more than it did for non-poor households. E-commerce participation did little to increase poor households’ future income growth expectation. Like many other poverty reduction programs, pro-poor e-commerce helps poor households with policy preferences but have yet to help them foster skills to prosper in the long run. The sustainability and quality of perceived relative economic gain for poor households are yet to be further observed and examined. All poverty reduction initiatives including pro-poor e-commerce must help poor households develop endogenous growth momentum to prosper beyond the effects of short-term pro-poor policies.
Keywords:
王瑜
E-commerce participation, sense of economic gain, poor households, nonpoor households
JEL Classification Codes: D19, Q13, O12
DOI: 1 0.19602/j .chinaeconomist.2020.05.08
1. Introduction
With the increasing penetration of internet applications, there has been a growing interest in the
social and economic effects of the internet among the public and academia. By breaking through market segmentation and broadening market access, e-commerce has emerged as a new channel for reducing poverty. Thriving e-commerce in China offers new experience for unraveling the effects of internet applications. Some studies suggest that e-commerce participation significantly boosts farmer households’ incomes (Lu and Liao, 2016; Zeng, et al., 2018). Such income growth stems from the falling price of perishable farm produce thanks to effective information supply (Xu, et al., 2013). E-commerce allows professional farmer households to earn a significantly higher income by increasing profit margin and sales (Zeng, et al. 2018).
Yet most samples employed in existing studies are collected from “Taobao villages” where e- commerce merchants flourish on Taobao, China’s largest e- commerce platform ( Zeng, 2018), e-commerce hotspot regions (Lu and Liao, 2016) or specific agricultural sectors (Zeng, et al., 2018; Xu et al., 2013). As such, their research conclusions may not apply to average rural households, especially poor households, in ordinary rural regions. Despite the growing public interest in recent years, few empirical studies have been carried out to examine the poverty reduction effects of e-commerce. Existing discussions on this topic are focused on the basic concepts and models. The extent to which e-commerce delivers economic opportunities to participants is yet to be examined, and is of great relevance to China’s “people-centered” development and the commitment to give people a “sense of gain.” Hence, this paper aims to reveal rural households’ sense of economic gain from e-commerce participation and whether such economic gains differ between poor and non-poor households.
With “sense of economic gain” as an outcome variable, this paper measures the inclusiveness of e-commerce participation by the following indicators, including “Self-evaluated income level relative to other households in the village,” and “Percentage of expected household income increase/decrease in 2018 over 2017.” Based on the nationwide village and household surveys conducted by the China Institute for Rural Studies (CIRS) at Tsinghua University in the summer of 2017, this paper identifies 6,242 rural households who have answered all questions in the surveys (together with their village conditions) to evaluate the effects of e-commerce participation on sense of economic gain. Compared with existing studies, this paper offers the following contributions: (i) It has extended the scope of research on the effects of e-commerce from specialized e-commerce villages to ordinary villagers and from professional farmer households to ordinary farmer households, including poor and non-poor households; (ii) it offers the first evaluation of the perceived economic benefits to e-commerce merchants in terms of relative income growth and future income expectation.
2. Literature Review and Research Hypotheses 2.1 Sense of Economic Gain and Determinants
Sense of economic gain is the primary outcome variable that this paper is concerned with. Research on the definition and indicators of the “sense of gain” among the populace remains limited, but may still offers some inspirations and support to this study. Existing studies define the “sense of gain” as an actual improvement in people’s living standards and subjective satisfaction (Yang and Zhang, 2019; Wen and Liu, 2018). Some academics regard the “sense of gain” as a multidimensional concept encompassing the sense of economic gain, i.e. an individual’s subjective level of satisfaction based on his/her real economic income (Yang and Zhang, 2019; Wen and Liu, 2018). Yang et al. (2019) classifies sense of economic gain into perceived income status relative to others, perceived income status compared with one’s past economic status, and expected future income growth and its realization (Yang and Zhang, 2019). Obviously, existing studies all examine sense of economic gain with a multidimensional approach. Moreover, Liang (2018) examines the sense of economic gain of low-income households from overall and relative dimensions. Based on the multidimensionality and data availability of sense of
economic gain, this paper measures sense of economic gain from two dimensions - self-evaluated level of household income relative to fellow villagers and future income growth expectation.
Sense of economic gain is subject to social environment, actual and perceived social status, and social policies. Ostensibly a subjective perception, the sense of gain is largely determined by certain objective factors (Zhang, 2018). Based on a national survey for low-income households conducted in 2016, a study finds that low-income households are less satisfied about the overall level of gain both in absolute and relative terms: External factors like region, community and the level of local economic development influence the sense of economic gain of low- income households both directly and indirectly through mediating effects (Liang, 2018). The focus of discussion in this paper is to unravel how e-commerce - a policy-supported industry with social and economic spillover effects - contributes to sense of economic gain among various groups of people in China.
2.2 Will E-Commerce Increase Sense of Economic Gain?
Via the internet as a new resource allocation mechanism (He, 2018), e-commerce allows farmers to earn a higher income by doing away with costly distribution links. In China, farm produce distribution is dominated by the wholesale market where price markups are applied at each level of distribution. Under this model, farmers wield no pricing power beyond their local market where farm produce is collected for distribution elsewhere, and cannot access the consumer market directly (Chen et al., 2019). Compared with the initial price quoted by farmers, farm produce ends up many times more expensive when they reach consumers after numerous distribution links, each with a price markup (Pan et al., 2018). Traditional resource allocation mechanism based on price signal may regulate the supply and demand of goods and services, but cannot rid the market of intermediaries the way resources are allocated over the internet (He, 2018). For farmers, these barriers will inevitably deprive them of economic gains that would otherwise come their way.
In contrast, internet applications optimize resource allocation by linking sellers with buyers across geographical barriers and allowing them to trade goods and services without resorting to an intermediary (He, 2018). With its information aggregation effects, the internet will create economic gains, efficiencies and social welfare beyond traditional economies of scale (Zhang, 2016). Empirical studies on Taobao villages and agricultural e- commerce platforms suggest that e- commerce is income enhancing for merchants (Lu and Liao, 2016; Zeng, et al., 2018).
Given the growing penetration of e-commerce and its potentials to upend the existing farm produce distribution market, the economic benefits of e-commerce participation should be universal for all farmer households at least in theory. Based on the above theoretical analysis, farmer households stand to gain from more efficient resource allocation through e-commerce participation. Hence, this paper puts forward the first hypotheses:
Hypothesis 1: E-commerce participation will increase farmer households’ sense of economic gain. Hypothesis 1a: E-commerce participation will increase farmer households’ sense of economic gain compared with their fellow villagers who did not participate in e-commerce.
Hypothesis 1b: E- commerce participation will increase farmer households’ income growth expectation.
2.3 Differences in Economic Gain from E-Commerce Participation among Farmer Households
According to existing theories, the lack of capital (Nurkse, 1953), especially human capital (Schultz, 1971), is the root cause of poverty. Poverty stems more from a dearth of capacity than from paltry incomes (Sen, 2001). Scant financial, human and social capital prevents the poor from economic and social participation. As proven in Chinese experience, human capital such as education is to blame as the chief culprit for yawning income gaps among rural households (Gao and Yao, 2006), and the case for capacity building among the poor is stronger than ever (Du, Park and Wang, 2005). As shown in
provincial panel data, e-commerce is more efficient at reducing poverty in regions with higher levels of human capital (Tang, et al., 2018). When evaluating the income effects of e-commerce participation, it is vital to control for differences in the endowment of poor and non-poor households for e-commerce participation. Even if such endowment differences are controlled for, there may still be systematic differences in the economic gains for poor and non-poor households from e-commerce participation.
Existing empirical studies have revealed how poor and non-poor households benefit differently from certain pro-poor programs. Based on households and village-level panel data of 2001-2004 and the matching method, Park and Wang (2010) finds that poverty reduction programs implemented for whole villages led to significantly higher incomes and consumption of prosperous households without benefiting the poor. Similar to e-commerce, cooperatives have also been regarded as an ideal vehicle for lifting the poor out of poverty through self-assistance and mutual assistance. Yet as Hu’s (2014) empirical study uncovers, high-income households benefited much more from Farmer Specialized cooperatives in poor regions than did poor households hamstrung by scant per capita assets to gain more from cooperatives.
Given the existence of capital constraints and empirical experience in similar sectors, there may be significant differences in economic gains from e-commerce between poor and non-poor households. Therefore, this paper puts forward the second group of hypotheses:
Hypothesis 2: Poor and non-poor households benefit differently from e-commerce participation; Hypothesis 2a: Relative economic gains from e-commerce are smaller for poor households than for non-poor households;
Hypothesis 2b: Poor households expect a smaller future income growth from e-commerce compared with non-poor households.
3. Data Source and Methodology 3.1 Data Source
The China Institute for Rural Studies (CIRS) at Tsinghua University conducted summer surveys on agricultural and rural development (CIRS Survey) for seven years from 2012 to 2018. This paper employs CIRS2017 data with the theme of “Rural Entrepreneurship and New Rural Business Models” collected from questionnaires and interviews at village and household levels. In addition to basic village and household information, questionnaires about villages and households also include such information as family-operated bed and breakfasts, e-commerce economy, rural entrepreneurship, and targeted poverty reduction.
This survey employs non-probability sampling1, including a combination of judgement sampling (a.k.a. expert choice or purposive sampling) and convenience sampling (a.k.a. accidental sampling): At the level of survey points (counties, townships and villages), the survey was carried out primarily with judgement sampling, and the CIRS expert team identified the topics and locations consistent with the
2 theme of the survey; at the level of household survey in selected villages, interviews were carried out
with households relevant with the theme of the survey. In June 2017, the CIRS expert team delivered lectures and trainings to interviewers. In July and August 2017, survey teams were dispatched to various villages.
3.2 Methodology Selection
The main purpose of this paper is to evaluate the effects of e-commerce participation on rural households’ sense of economic gain. For two reasons, the propensity score matching (PSM) appears to be the most appropriate method. First, the CIRS2017 data employed in this paper are not probability sampling data, and sample matching is an adjustment method for resolving the problem of statistical inference from non-probability sampling (Jin and Liu, 2016). The PSM is widely used in non-probability sampling inference with good effects (Liu, 2018). Based on the PSM, a statistical inference can be carried out with Efron’s ( 1979) bootstrap repeated sampling technique using given observations without other assumptions or new observations. Second, there is a “selection bias” due to differences in households’ initial resource endowment. Since households decide on their own whether or not to participate in e-commerce, it is necessary for such self-selection to be treated. Also known as the Rubin Causal Model (RCM) (Holland, 1986), Rubin’s (1974) “counterfactual framework” evaluates the treatment effect with counterfactual characteristics as missing data. As a data balance method, the matching method identifies the members of a non-intervention group similar to those of the intervention group on the covariate, and uses the average result of the non-intervention group as a proxy to estimate the counterfactuals of the intervention group (Guo and Frazer, 2012). This paper employs Rosenbaum and Rubin’s (1983) PSM with “propensity value” as the distance function, which balances the covariate with selection bias to obtain a uniform distribution.
Study on the sense of economic gain from e-commerce participation can be seen as an evaluation of the “treatment effect” . Rural households involved in e-commerce business constitute the “treatment group”, and those not involved in e- commerce business are the “control group”. Referencing the econometric application of the counterfactual framework ( Chen, 2014), the average difference of outcome variable Yi (income level or income expectation) is subject to whether a household is involved in e-commerce, expressed as:
(1) In equation (1), i is the number of individual household. Dummy variable Di ={0,1} denotes whether individual household i is involved in e-commerce business (1=Yes; 0=No). The outcome variable (sense of economic gain) Y is subject to a group of explanatory variables X, the average of which is influenced by e-commerce participation D. (− Y1i Y0i) or is the average treatment effect (ATE) of e-commerce, and the average treatment effect on the treated (ATT) for rural households involved in e-commerce business is expressed as:
(2) Since some of the samples may not participate in e-commerce at all, a simple comparison of the outcome variable between participants and non-participants may give rise to a selection bias. Thus, ATE consists of ATT and selection bias. For officials and policymakers, ATT matters more since it measures participants’ gross return.
The reality is that households are either involved in e-commerce business or not involved at all. Namely, one of the choices made by households will always be observed. If a household is involved in e-commerce business, Y1i will be observed, but the potential result of non-participation cannot be observed. If a household is not involved in e-commerce, Y0i will be observed, but the potential outcome of participation cannot be observed. That is to say, the potential outcome of the counterfactual choice
is a missing value. The evaluation of the treatment effect in the observation data comes down to the treatment of missing data. Propensity value analysis has been proven to be an effective statistical method for evaluating the treatment effect based on observation data. PSM identifies an individual j of the control group who corresponds to an individual i of the treatment group, whose measurable covariates are similar based on parametric or non-parametric regression (this paper employs logit model to estimate the propensity value), so that the outcome variable of individual j can be used as the counterfactual reference for individual i.
Based on the sample calculation treatment method after PSM, we proceed to estimate the ATT of rural households involved in e-commerce business with the following equation:
3.3 Variable Explanation and Statistical Characteristics
In this paper, the explained variable reflects rural households’ sense of economic gain from different dimensions. Existing empirical research recognizes the viability of measuring perceived gain among the Chinese public on both dimensions of time and reference group (Lyu and Huang, 2018). Similarly, this paper carries out an analysis on both dimensions of comparison with peers and future income expectation from the CIRS2017 questionnaire as variables of sense of economic gain. The outcome variable of perceived income gain compared with peers is “Self-evaluated income level compared with fellow villagers” at the time of survey (summer of 2017), which is divided into five grades from low, below average, average, above average to high. This question asks households about their perceived relative income level in the village. Future income expectation is measured by “Percentage of expected household income growth (reduction) in 2018 over 2017.”
To ensure the use of the same samples for analysis on different dimensions, this paper retains questionnaires with answers to all questions, including 6,242 households. Among them, 13.8% (859) of all rural households are involved in e-commerce business; 33.5% of (2,093) rural households are registered poor households, and the rest (4,149) are non-poor households; 8.1% of poor households and 16.6% of non-poor households are involved in e-commerce.
Variables that may affect households’ sense of gain include community environment, human capital, and material capital. These variables are based on 2016 information, which precedes perceived gain variable evaluated during the survey of 2017 and meets the above-mentioned criteria. The descriptive statistics of relevant variables are shown in Table 1. The t-test of mean difference suggests that apart from the household head’s level of education, significant differences exist in the explained variable and covariates between households involved in e-commerce and those not involved. Based on the definitions of variables, households not involved in e-commerce are significantly disadvantageous to those that are involved in terms of endowment and external environment. To overcome the self-selection bias of households with respect to e-commerce participation, it is highly necessary to adopt an ATT evaluation model.
4. Analysis of Empirical Results
(3)
4.1 Measurement Results of the Sense of Economic Gain from E-Commerce Participation
Table 2 and 3 identify the treatment effects of sense of economic gain from e- commerce participation on the two dimensions. Given the existence of numerous comparable control group samples and the robustness of results, this paper simultaneously employs k-nearest neighbor matching and kernel matching methods, and calculates standard error with bootstrapping method in reporting estimation